-
Banned
I was surprised to see people even as PC smart as Bela say things like, the GT doesn't use much power. I want you guys to think about this for a minute...
What physical force is acting on the board to heat it?
Answer is = electron flow measured in watts of power!
You gottum other answer, I'd love to hear it from you paleface! To say a board heats, but uses little power is complete and utter nonsense. If a power supply can't give the necessary wattage at a certain voltage, it will try to make it up in amps. Remember, amps X volts = watts. It's simple math and no belief system will overcome math. If the wattage can't be delivered at a certain voltage by the power supply, it will try to make it up in amperage. Being fed too many amps at a lower voltage VS. less amps at higher voltage to give the needed wattage = power to run the board, causes the card to heat beyond its designed limits.
I am NOT, repeat, NOT saying that this is the only or the most important concern in all matters. You still have to attend to case cooling and cooling of the card itself under even the best of power deliveries.
However, believing your heating situations don't have a power component to them is just that, a belief system.
-
Senior Member
Talking of paleface...
Hey HEY! In the context I told that, it was in comparison to the older nvidia 5900 series wich many have been using for more than 2 years with much less suited PSUs (for a laugh, my old one would recommend a 250W on it's box). Power consumption in a GT is around 55W while 5900 ultras and 5950 are in the 60-80W and up range as with 6800 ultras.
When compared to people running old 4200's or 9500 it's quite a huge step, but not compared with the last 2 years's top end, as most here came from 9800s. These had similar power consumption, but WAY less transistors and less core area, less power regulation on mosfets due to bigger processing needing more v, and hence less heat directly on core and pcb. Still, the strain on the psu was the same and where running (supposedly) fine. That was what I meant in other post, everyone knows 6800s are power hogs no matter what, so please take that comment with a grain of salt. 
http://www.xbitlabs.com/articles/vid...-power_10.html one of many articles about this.
Best regards!
edit: small edit
-
Banned
No worries Bela, you forget more about computers in a day than I will probably learn in a lifetime!
Still, you take even the PCI Express versions of the 6600 cards for example... Watts of electrical power used get expressed in light, electro-mechanical motion like fan motors or heat. This directly translates into a certain amount of calories of heat. The calories of heat being produced by even the 6600 cards shows the manufacturers are under-reporting the actual power usage of the boards in question or the temps would be lower, especially with the heat sinks and fans carrying away the heat. I do not know if the manufacturers under-reporting the power consumption needs of the parts in question is willfull or just a mistake, but I suspect the former.
You bring up another excellent point about PCI Express. They get all their power through the PCI buss and as a result, are limited by whatever power your particular MOSFET/motherboard power circuitry setup can deliver. At this point, I doubt the boards are truly up to giving rock steady performance, especially in dual card modes. Trying to beat the information out of the motherboard manufacturer as to how well their part will deliver the needed power saturation to the PCI buss seems next to impossible right now. I predict a whole series of picky, transient power related failures for most of the boards being produced today. First the gamers will have to wake up to what is happening and then pressure will have to be put on the manufacturers by the consumer before this problem ever gets fixed, so you guessed it, a 1-2 year process if it ever gets fixed at all.
Ah yes, I pray for the day that motherboards will be sold based on their specifications vs. cool names!
-
Member
Are you saying that Mobo's are power starved at present, or that they will be in the near future?
-
Senior Member
Sorry for the OT on the temps, but following the last 2 posts:
Every mobo is starved right on boot up
, this is where it draws almost infinite current from your psu and all components are filling those caps etc, even tough it's for a very short period it gives a lot of strain because the psu hasn't time to stabilize.
As an example, probably the most extreme mobo around is the dfi nf4, where IIRC they state a 94a peak on 12v right on boot up
. People are having a lot of trouble with old psus on it or even small rated dual rail ones where the cpu line is rated less than 20a, and some brand 430w can't even make it boot up. Just go to dfi-street.com and see how many long threads about PSUs you can find in there!
For PCIe starvation, well, cards are going the old route again, current PCIe monsters already need extra plugs. The aforementioned board needs an extra molex plugged to it, the 24 pin, the 4 pin, and even a small fd power plug (and I guess something more that I'm forgetting right now!). With many "weaker" but equal power sources around , it's easier to make them more stable across the whole board. Currently it's working pretty fine given the crazy clocks people are getting on it.
LB, for power rating, manufacturers should make requirements from all the power a component draws right on the plugs or sockets, not matter if chips use it, or it's dissipated on a power regulation section. Still, drawing a lot of 12v to dissipate on a 1.x v feed is a huge waste. Add this to less than optimal PSU efficiency, under light load most PSUs are about 60% efficient and the best ones (you'd be surprised on how many brand psus would do less than 70% on older models) will be at most 80% on full load which is a big waste too.
"The horror, the horror..."